About:

Onur Solmaz is an engineer and researcher with interests in languages, math, and sports, contributing to Manim Community and working at TextCortex.

Website:

Specializations:

Interests:

Languages Math Physics Thinking Strength training Sports

Outgoing Links:

Subscribe to RSS:
The author recounts the journey of developing a coding agent before ChatGPT, highlighting the evolution of AI coding tools and their functionalities.
GitHub must fundamentally alter its trust and data models to safely accommodate agentic workflows and prevent potential damage from bots and agents.
The post advocates for 'AI hygiene' rules to prevent emotional attachment and psychological issues stemming from AI interactions.
The author shares their transformative experience using Claude Code, a tool that has significantly enhanced their productivity as a developer. They describe how the tool has allowed them to merge repositories, implement new featur...
Agentic coding tools should enhance user control by implementing flexible message queuing options to improve productivity and streamline workflows.
The future of software development involves managing multiple AI agents simultaneously, transforming attention management and the nature of knowledge work.
OpenClaw's success is attributed to its integration of AI into popular messaging apps, enhancing user experience for developers through familiar interfaces.
A new CLI flag convention, --skill, is proposed to improve skill distribution for AI agents, emphasizing the need for better documentation and usability.
AI can significantly speed up coding tasks, but deep understanding and adaptive learning remain crucial for success in technical fields.
The author discusses recent enhancements made to their blog, including features like light/dark mode, permalinks, and content copying. They reflect on their past procrastination regarding design, emphasizing a shift in perspective...
The post discusses the migration from CLAUDE.md to the AGENTS.md standard, which is an open standard for providing context to AI agents in codebases. It criticizes the CLAUDE.md naming convention and outlines the steps for migrati...
The 'Christmas of Agents' offers a unique opportunity for immersive coding projects, as demonstrated by the author's extensive development work during the holiday season.
OpenClaw's latest update enables persistent integration of Claude Code in Discord and Telegram, eliminating account ban risks and enhancing user experience.
OpenClaw's latest update enables persistent integration of Claude Code in Telegram and Discord, enhancing user experience without the risk of account bans.
LLM-based tools are insecure if they access private data, while open-source software is more secure due to community involvement in fixing vulnerabilities.
The post argues that LLM-based tools pose security risks and that open-source software is more secure than closed-source due to community involvement in fixing vulnerabilities.
A secure workflow for managing coding agents in GitHub is proposed, highlighting the need for first-class support for agents in the platform's trust model.
Anthropic's non-linear pricing strategy for its API is deemed ineffective compared to OpenAI's, potentially limiting its market competitiveness and boosting open-source alternatives.
The post highlights the benefits of a monorepo structure for developers and shares insights on using AI to create personal tools efficiently.
A solution for managing AI-generated documentation in codebases, introducing the SimpleDoc specification for organized and timestamped files.
This blog post adapts Google's Code Review Guidelines to GitHub terminology, aiming to make the guidelines more accessible. It outlines the code review process, what reviewers should look for (design, functionality, complexity, te...
The author discusses how their programming habits have evolved over the past decade, particularly since the introduction of Claude Code. They now prefer using typed, compiled languages like TypeScript, Rust, and Go over Python for...
The blog post discusses the challenges faced when using the 'uv' Python package manager, particularly with the AI model Claude Code, which often ignores instructions to use 'uv run python' instead of standard Python commands. The ...
Sholto Douglas and Trenton Bricken discuss the advancements in AI and predict what tasks will be solved by next year. They believe that any white-collar task can be automated if the labs put resources into it, and they predict whi...